mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-10-22 12:33:33 +02:00
26beaec6ae
We still need to add constant folding of vector comparisons to fold the tests for targets that don't support the respective min/max nodes I needed to update 2011-12-06-AVXVectorExtractCombine to load a vector instead of using a constant vector to prevent it folding Differential Revision: http://reviews.llvm.org/D12118 llvm-svn: 245503
19 lines
807 B
LLVM
19 lines
807 B
LLVM
; RUN: llc < %s -mtriple=x86_64-apple-darwin -mattr=+avx | FileCheck %s
|
|
; PR11494
|
|
|
|
define void @test(<4 x i32>* nocapture %p) nounwind {
|
|
; CHECK-LABEL: test:
|
|
; CHECK: vpxor %xmm0, %xmm0, %xmm0
|
|
; CHECK-NEXT: vpmaxsd (%rdi), %xmm0, %xmm0
|
|
; CHECK-NEXT: vmovdqu %xmm0, (%rdi)
|
|
; CHECK-NEXT: ret
|
|
%a = load <4 x i32>, <4 x i32>* %p, align 1
|
|
%b = call <4 x i32> @llvm.x86.sse41.pmaxsd(<4 x i32> %a, <4 x i32> zeroinitializer) nounwind
|
|
%c = shufflevector <4 x i32> %b, <4 x i32> undef, <8 x i32> <i32 undef, i32 undef, i32 undef, i32 undef, i32 0, i32 1, i32 2, i32 3>
|
|
%d = shufflevector <8 x i32> %c, <8 x i32> undef, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
|
|
store <4 x i32> %d, <4 x i32>* %p, align 1
|
|
ret void
|
|
}
|
|
|
|
declare <4 x i32> @llvm.x86.sse41.pmaxsd(<4 x i32>, <4 x i32>) nounwind readnone
|